Emotions are critical for action selection in both humans and robots.

It is tempting to think that in a complicated situation pure reason will enable one to choose the best course of action. Reasoning is unquestionably very important. But significant work, eg Damasio, has shown that on its own reasoning often leads to poor decision-making. The missing ingredient is emotion. Arbib’s and Fellous’ paper makes the same point. Intriguingly, though, they then propose that emotions are systems that are substrate-indifferent. Thus, they can be instantiated in AIs or robots too. Still more intriguingly, they suggest that in a complex environment a robot with emotions would do better than a robot without. The former would prioritise its next steps more effectively than the latter. This is a rich seam to mine en route to the development of artificial general intelligence.

Link to paper: https://www.ncbi.nlm.nih.gov/pubmed/15556025

You may also like to browse other AI academic papers: https://www.thesentientrobot.com/category/ai/ai-academic-papers/